Sensitivity Analysis with Control Chart
Sensitivity Analysis Integrated with Statistical Process Control Charts · Also known as: SA-SPC integration, control chart sensitivity analysis, SPC sensitivity assessment, sensitivity-enhanced control charting
Sensitivity analysis integrated with control charting evaluates how uncertain or varying inputs — such as sample size, subgroup frequency, distribution assumptions, or measurement error — affect the detection performance of a statistical process control chart. By quantifying which parameters most strongly influence chart metrics such as the average run length (ARL) or false alarm rate, engineers can design more robust monitoring schemes and understand where control chart conclusions are fragile.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use this approach when deploying a control chart in a setting where process parameters (mean, variance) are estimated from limited baseline data, the underlying distribution may deviate from normality, measurement uncertainty is non-negligible, or sampling constraints force a design trade-off. It is especially valuable before committing to a long-term monitoring scheme in high-stakes processes (pharmaceutical manufacturing, semiconductor fabrication, safety-critical systems). Do not apply it when a process is already well-characterised with abundant historical data and the distributional assumptions are firmly validated — in that case standard control chart design is sufficient without the overhead of sensitivity analysis.
Strengths & limitations
- Reveals which chart design choices are truly influential versus which can safely follow convention, enabling focused investment in parameter estimation.
- Prevents overconfidence in chart performance figures derived under idealized assumptions that may not hold in practice.
- Compatible with all major chart types (Shewhart, CUSUM, EWMA, multivariate T2) and with both simulated and analytical ARL models.
- Supports risk-informed decision-making by quantifying the uncertainty band around advertised ARL performance.
- Improves communication with process owners by translating statistical assumptions into practical impact on detection speed.
- Requires a mathematical or simulation model of chart performance, which adds complexity beyond standard chart implementation.
- Variance-based global sensitivity analysis (Sobol) is computationally intensive — many thousands of model evaluations may be needed for high-dimensional parameter spaces.
- Results are only as credible as the distributional assumptions placed on the uncertain inputs; poorly specified input ranges can mislead.
- Does not directly redesign the chart — it diagnoses fragility but the engineer must translate findings into revised design decisions.
Frequently asked
What sensitivity analysis method should I use with a control chart model?
For an initial screening with many parameters, the Morris elementary effects method is computationally efficient and identifies non-influential inputs to eliminate. For a thorough ranking with quantified contributions, Sobol variance-based indices (first-order and total-effect) are the gold standard. Local OAT perturbation is fast but misses interactions and is not recommended as the sole method.
What is ARL and why is it the key performance metric?
ARL (average run length) is the expected number of samples taken before a control chart signals. ARL0 refers to the in-control state — a large ARL0 means few false alarms. ARL1 refers to an out-of-control state — a small ARL1 means the chart detects problems quickly. Sensitivity analysis of these two quantities tells you how stable these desirable properties are under realistic variation in chart parameters.
Can this approach be applied to multivariate control charts?
Yes. Hotelling T2 charts, MEWMA, and MCUSUM charts have additional parameters (covariance matrix estimation, number of variables) and their ARL properties are even more sensitive to distributional and estimation assumptions than univariate charts. Global sensitivity analysis is particularly valuable in multivariate settings where analytical ARL expressions are unavailable and simulation is the primary tool.
Does sensitivity analysis replace Phase I chart design?
No — it complements it. Phase I analysis uses historical data to estimate control limits. Sensitivity analysis then asks: if those estimates are off by a plausible amount, how much does chart performance change? The two steps together produce a more reliable and transparent deployment than Phase I alone.
How many simulations are needed for a reliable Sobol index?
A common rule of thumb is N*(k+2) model evaluations, where N is a base sample size (often 1000–10000) and k is the number of uncertain inputs. For a chart model with 5 uncertain inputs and N=2000, this means around 14000 simulations — feasible for most ARL simulation models within minutes on a modern workstation.
Sources
- Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana, M., & Tarantola, S. (2008). Global Sensitivity Analysis: The Primer. Wiley. ISBN: 978-0470059975
- Montgomery, D. C. (2020). Introduction to Statistical Quality Control (8th ed.). Wiley. ISBN: 978-1119399308
How to cite this page
ScholarGate. (2026, June 3). Sensitivity Analysis Integrated with Statistical Process Control Charts. ScholarGate. https://scholargate.app/en/experimental-design/sensitivity-analysis-with-control-chart
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Control chartExperimental design↔ compare
- Design of experimentsExperimental design↔ compare
- Robust Control ChartExperimental design↔ compare
- Sensitivity Analysis with Process Capability AnalysisExperimental design↔ compare
- Sensitivity analysis-integrated design of experimentsExperimental design↔ compare
- Statistical Process ControlExperimental design↔ compare